Cross-language perception of Mandarin affricates in conversational vs careful speech.

Abstract

Two affricates “zh” and “j” in Beijing Mandarin are often confused by English speakers learning the language [Dow 1972]. An experiment based on the work of Logan in 1989 training Japanese speakers to distinguish the difference between English /∈vr/ and /l/ was modified to include training tokens taken from native Mandarin speakers’ phone-conversational speech. “j” is described by Lee and Zee [2003] as laminal pre-palatal, and “zh” is described as apical post alveolar. In careful speech, the sounds are in complementary distribution, with “j” always followed by a high front vowel or a glide. A preliminary perception experiment shows whether native Mandarin speakers are able to distinguish between the pair of affricates simply by hearing the affricate alone, or if they are only able to distinguish them when hearing the vowel information as well. In conversational speech, the following high front vowel may be deleted, so natives’ perception of the two affricates may differ for casual speech. English speakers with no Mandarin experience also participate in a perception task to show the effects of training with conversational speech tokens. This study elucidates how learners acquire categories despite the variability of natural speech.